docs: sync roadmap after core coverage

[salim4n]

Summary

• Mark PR test: add core package coverage #44 as completed in the backlog.
• Update roadmap, milestones, post-feat: add bandit strategy prototype #20 roadmap and alpha readiness after environment episodes, trajectory reports and core tests.
• Add the next open-source sequence: CLI ergonomics, deeper adapter examples, then dogfood-driven policy optimization.

Next planned work

1. PR feat: add CLI history baseline and regression ergonomics #46: CLI history, baseline selection and regression ergonomics.
2. PR feat: add ecosystem adapter examples #47: deeper mocked examples for LangChain, LangGraph, Mastra and Vercel AI SDK adapters.
3. PR feat(core): add runtime validation helpers #48: lightweight policy optimization loop, blocked until useful dogfood trajectories exist.

Out of scope

• Runtime source changes
• Package API changes
• Test implementation changes
• npm publication
• IgnitionRAG repo work
• PPO / GRPO training / model fine-tuning

Validation

• bun run lint
• git diff --check

Note: packages/*/dist was cleaned before lint because generated build output is ignored but Biome scans it when present.